Emmanuel Bourg wrote on Friday, March 05, 2004 11:38 AM:
> J�rg Schaible wrote:
>
>> after recognizing, that you've attached the diff, I had a look at the
>> changes in the test case. I am not quite sure, that this is really
>> good. Intentionally I thought, you would return an empty
>> configuration only, if the pattern equals the key, but you do so now
>> also if the pattern does not match at all.
>>
>> This will break my usage of Configuration for certain and I am
>> probably not alone here.
>
> You expect subset() to return a null configuration instead of
> an empty
> configuration ?
>
> I see one issue with the current implementation, it is not
> possible to
> access the property in the parent configuration with a key
> matching the
> prefix of the subset (the subset root property):
>
> conf.setProperty("x.y.z", "1");
> subset = conf.subset("x.y.z");
>
> Here subset is empty and I can't read the value of "x.y.z".
Well, in the previous implementation this was thew special case and the subset has
exactly one property "x.y.z" with value 1. Your implementation delivers an empty
configuration, right? Fine with me.
Looking at the changes for teh test cases, I have the impression that
conf.setProperty("x.y.z", "1");
subset = conf.subset("x.y.z.not_there");
will also return an empty configuration, whereas this return value was previously
always just null.
> If I call
> subset.getProperty("") it will return null because the key "x.y.z."
> doesn't exist in the parent configuration. I think I'll change the
> implementation I suggested to allow the use of a null (or "" ?) key to
> access the subset root property, that's
> subset.getProperty(null) returns
> "1".
OK. this makes more sense than the empty configuration.
Regards,
J�rg
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
